Article pubs.acs.org/jmc
Identification of Small Molecule Inhibitors of Amyloid β‑Induced Neuronal Apoptosis Acting through the Imidazoline I2 Receptor Marisol Montolio,† Elisabet Gregori-Puigjané,‡ David Pineda,† Jordi Mestres,*,‡ and Pilar Navarro*,† †
Cancer Research Program and ‡Biomedical Informatics Program, IMIMHospital del Mar Research Institute and University Pompeu Fabra, Parc de Recerca Biomèdica (PRBB), Doctor Aiguader 88, 08003 Barcelona, Catalonia, Spain S Supporting Information *
ABSTRACT: Aberrant activation of signaling pathways plays a pivotal role in central nervous system disorders, such as Alzheimer's disease (AD). Using a combination of virtual screening and experimental testing, novel small molecule inhibitors of tPA-mediated extracellular signal-regulated kinase (Erk)1/2 activation were identified that provide higher levels of neuroprotection from Aβ-induced apoptosis than Memantine, the most recently FDA-approved drug for AD treatment. Subsequent target deconvolution efforts revealed that they all share low micromolar affinity for the imidazoline I2 receptor, while being devoid of any significant affinity to a list of AD-relevant targets, including the N-methyl-D-aspartate receptor (NMDAR), acetylcholinesterase (AChE), and monoamine oxidase B (MAO-B). Targeting the imidazoline I2 receptor emerges as a new mechanism of action to inhibit tPA-induced signaling in neurons for the treatment of AD and other neurodegenerative diseases.
■
INTRODUCTION Cell signaling pathways are essential to regulate cellular activity. They are responsible for key cell functions such as growth, metabolism, cytoskeletal regulation, adhesion, translational control, DNA damage, and apoptosis. Well-orchestrated activation of cell signaling pathways is required during cell development and differentiation as well as for the maintenance of cellular homeostasis. Among the large number of intracellular signaling pathways, the extracellular signal-regulated kinase (Erk) is one of the first and best-characterized signal transduction networks. The Erk1/2 kinases are serine/ threonine kinases that belong to the mitogen-activated protein kinase (MAPK) superfamily of signaling proteins, and they integrate cellular signals in response to a variety of stimuli that affect cell proliferation, differentiation, and survival. Altered activation of Erk signaling pathway results in cell dysfunctions that have been found to be linked to severe pathologies such as cancer or neurological disorders.1,2 In the central nervous system (CNS), induction of Erk cascade is critical for learning and memory,3 and aberrant Erk activation has been particularly reported in the pathogenesis of Alzheimer's disease (AD),4−6 a devastating CNS pathology characterized by progressive neuronal loss accompanied by cognitive decline and dementia.7 Amyloid-β (Aβ) is recognized to play a central role in the synaptic dysfunction and neurodegeneration linked to AD,8 and its accumulation is known to induce sustained Erk activation both in vitro5,9 and in vivo,5,10 ultimately leading to neuronal death. However, the molecular mechanisms underlying Erk stimulation by Aβ are not yet completely understood. In this respect, it was previously reported that tissue plasminogen activator (tPA) is involved in this event.6 Even though tPA is mainly known as a fibrinolytic agent in the conversion of plasminogen into plasmin,11 a recent © 2012 American Chemical Society
study showed that it mediates Aβ-induced apoptosis through a rapid and sustained catalytic-independent activation of the Erk1/2 signal transduction pathway.6 Indeed, Aβ treatment after pharmacological inhibition of tPA or in primary neurons obtained from tPA−/− mice failed to induce Erk activation and neuronal apoptosis,6 indicating that inhibition of tPA-induced signaling in neurons could represent a novel approach to prevent Aβ-mediated neurotoxicity and thus become a valuable strategy in the treatment of AD. On this basis, the present study aims at probing the link between tPA-mediated Erk1/2 activation and Aβ-induced apoptosis with small molecules as a means to identify protein targets that could play a key role in managing this connection. Using MK-801, a previous work confirmed that the NMDA receptor (NMDAR) is one of those targets.6 Through the identification of novel small molecules that inhibit the Erk1/2 signaling pathway and protect against Aβ neurotoxicity, this study reveals that the imidazoline I2 receptor is another one of those targets, paving the way to novel strategies to AD pathogenesis.
■
RESULTS Novel Inhibitors of tPA-Induced Erk1/2 Activation. Identification of small molecules inhibiting tPA-induced Erk1/2 activation was performed in primary cultures of mouse hippocampal neurons, as this cellular system can mimic the in vivo situation during AD, where hippocampus is one of the most affected regions. Given the limited capacity of an experimental assay involving the use of primary cultures, we had to revert to computational approaches to prioritize a small Received: July 19, 2012 Published: October 25, 2012 9838
dx.doi.org/10.1021/jm301055g | J. Med. Chem. 2012, 55, 9838−9846
Journal of Medicinal Chemistry
Article
Figure 1. Chemical structures of the two reference compounds, Memantine and MK-801, and the 17 compounds purchased from commercial providers and analyzed in this work.
Figure 2. Inhibition of tPA-induced neuronal Erk1/2 activation. Primary mouse hippocampal neurons were treated with medium alone (−, negative control), tPA (20 μg/mL, 1 h), or tPA in the presence of Memantine, MK-801, or molecule 1−17 at 30 μM. Phospho- (activated) and total-Erk1/2 were detected by ELISA. The graph corresponds to the quantification of Erk1/2 activation with respect to tPA levels, normalized to the total amount of Erk1/2. Statistical analyses were performed with a two-tailed Student's test (n = 4), and data are mean values ± SEMs. *P ≤ 0.05 and ** P ≤ 0.001, as compared to tPA conditions.
number of selected compounds that would ultimately go into testing. As mentioned above, a previous study showed that MK801, a high-affinity noncompetitive NMDAR antagonist,12 was able to abolish completely the activation of the Erk1/2 signaling pathway by tPA.6 The same effect was later observed with Memantine (vide infra), a low-affinity noncompetitive NMDAR antagonist13 currently used widely as first-line
treatment of AD. Because information on two small molecule inhibitors of tPA-mediated Erk1/2 activation was available, similarity-based virtual screening emerged as an obvious option to prioritize compounds for experimental testing. In this respect, we used a similarity approach based on Shannon entropy descriptors (SHED)14 derived from topological feature distributions that were successfully applied recently in the 9839
dx.doi.org/10.1021/jm301055g | J. Med. Chem. 2012, 55, 9838−9846
Journal of Medicinal Chemistry
Article
Figure 3. Inhibition of Aβ-induced neuronal apoptosis. Primary mouse hippocampal neurons were treated with Aβ (20 μM, 72 h) alone or in the presence of Memantine or different molecules (4−7, 9, 10, 13, 16, or 17) at 30 μM. Neuronal apoptosis was detected by TUNEL staining. The graph represents the percentage of apoptotic cells relative to the total number of nuclei and referred to Aβ conditions (100%). Neurons treated with medium alone (−) or DNase were used as negative and positive controls, respectively. Statistical analyses were performed with a two-tailed Student's test (n = 4), and data are mean values ± SEMs. *P ≤ 0.05 and ** P ≤ 0.001, as compared to Aβ conditions.
tested) showed similar (6, 9, 13, and 16) or enhanced (4, 5, 7, 10, and 17) levels of inhibition of Erk1/2 as compared to those obtained with Memantine or MK-801. All nine bioactive molecules identified have scaffold structures that are strictly different from the ones present in the two reference compounds (Figure 1). In terms of pharmacophoric distributions, they can be broadly organized in three clusters. The largest pharmacophoric cluster, cluster I, is composed of five out of the nine bioactive molecules (4, 5, 6, 7, and 13), the structures of which are characterized by the presence of two aromatic centers, each of them being two bonds away from a positively charged amine, the essential pharmacophoric signature of MK-801. A second pharmacophoric cluster, cluster II, is defined by two molecules (9 and 10) that differ from the structures present in cluster I by the fact that one of the aromatic centers is substituted by a second positively charged amine, a bioisosteric replacement used commonly in medicinal chemistry. The final cluster, cluster III, contains the two remaining bioactive molecules (16 and 17) that retain the main pharmacophoric features present in the structure of Memantine. These 11 molecules will be considered for further testing for their ability to inhibit Aβ-induced neuronal apoptosis. Inhibition of Aβ-Induced Neuronal Apoptosis. It has been previously reported that Aβ induces neuronal toxicity through Erk1/2 activation5,9,17 and that these effects are mediated by tPA.6 Accordingly, the next step was to test the nine molecules identified as inhibitors of tPA-mediated Erk1/2 activation for their neuroprotective effects against Aβ-induced apoptosis. Primary cultures of mouse hippocampal neurons were treated with aggregated Aβ alone (taken as positive control) or in the presence of each bioactive molecule, and apoptosis was quantitatively assessed by measuring the number of positive nuclei detected by TUNEL labeling. The percentages of Aβ-induced apoptosis achieved in the presence of the different small molecules in comparison to values observed after single Aβ treatment (normalized to 100%) are presented in Figure 3. Untreated cells were used to provide the basal level of apoptosis in primary cultures (10.5 ± 0.8%).
identification of novel compounds that reactivate latent HIV1.15 Accordingly, MK-801 and Memantine were taken as reference structures (Figure 1) from which similar compounds were searched in a commercial chemical catalogue composed of over 2 million compounds. A total of 1758 compounds were retrieved under a SHED Euclidean distance cutoff of 1.5. Of those, 1453 and 305 were found to have structures with topological pharmacophoric features distributed similarly to MK-801 and Memantine, respectively (Table S1 in the Supporting Information). To reduce the number of compounds further, a scaffold analysis was performed by extracting the atomic frameworks of all molecules using an internally developed chemical graph identifier:16 Four hundred forty-six scaffolds were identified, and a representative molecule of each scaffold was selected. Finally, a cost filter was applied to those 446 molecules. After visual inspection, 17 cost-effective molecules with scaffold topologies strictly different from MK801 and Memantine were prioritized for purchase from commercial providers (Table S2 in the Supporting Information). The structures of these molecules are collected in Figure 1, of which 14 (molecules 1−14) and 3 (molecules 15−17) come from the sets of molecules having pharmacophoric distributions similar to MK-801 and Memantine, respectively. The 17 molecules selected were experimentally tested for their capacity to inhibit Erk1/2 activation induced by tPA treatment of hippocampal neurons as a primary test for the more elaborate testing of Aβ-induced apoptosis. Detection of activated Erk1/2 was performed by ELISA using an antibody specific for phosphorylated Erk1/2, and values were normalized using an antibody against total Erk1/2 (Figure 2). The lower and upper Erk1/2 activation limits were established with the basal level observed with untreated cultures and its stimulation in the presence of tPA, respectively. Memantine and MK-801 were taken as positive controls, and their levels of inhibition of Erk1/2 activation were used as criteria for the identification of novel bioactive molecules. As can be observed, 10 molecules showed statistically significant inhibition of Erk1/2 activation by tPA, and nine of them (53% of all molecules purchased and 9840
dx.doi.org/10.1021/jm301055g | J. Med. Chem. 2012, 55, 9838−9846
Journal of Medicinal Chemistry
Article
Figure 4. Molecules 9, 10, and 17 show neuroprotection against Aβ toxicity. Primary mouse hippocampal neurons were treated with Aβ (20 μM, 72 h) (a−c) or Aβ in the presence of Memantine or different molecules (9, 10, and 17). Cells were analyzed by immunofluorescence microscopy to detect DNA fragmentation (TUNEL, green), total neuronal nuclei (NeuN, red), and overlay (merge). Treatment with DNase was used as a positive control. Bar, 50 μm.
statistical significance, apoptotic protection (5, 13, and 16). Accordingly, focus turned then into further characterizing the three molecules that showed the highest levels of neuroprotection from Aβ-induced apoptosis (9, 10, and 17). Figure 4 shows the efficiency of molecules 9, 10, and 17 to protect against apoptosis triggered by Aβ by double immunofluorescence of apoptotic nuclei (TUNEL, green) and total neuronal nuclei (NeuN, red), as well as the combined images (merge). Both DNase (Figure 4a−c) and Aβ (Figure 4d−f) treatments lead to extensive neuronal apoptosis, whereas pretreatment with Memantine (Figure 4g−i), 9 (Figure 4j−l), 10 (Figure 4m−o), or 17 (Figure 4p−r) rescue the Aβ pro-apoptotic
Neurons were also treated with DNase as an additional positive control of apoptosis (116.4 ± 18.3%). Finally, the reduction of cellular apoptosis observed when neurons were preincubated with Memantine prior to Aβ addition (42.3 ± 2.7%) served to establish the target criteria for apoptotic protection to be achieved by novel neuroprotective small molecules. The results show that out of the nine molecules selected in the first round of Erk-inhibition ELISA screening, five of them provided better levels of neuroprotection from Aβ-induced apoptosis than Memantine (4, 6, 9, 10, and 17); one also showed a significant decrease of Aβ-induced apoptosis (7), and another three were found to achieve only weak, not reaching 9841
dx.doi.org/10.1021/jm301055g | J. Med. Chem. 2012, 55, 9838−9846
Journal of Medicinal Chemistry
Article
Figure 5. Single concentration in vitro profile (% inhibition at 10 mM) of Memantine and compounds 9, 10, and 17 across a list of 19 targets (see the text for target acronyms). Dose−response curves corresponding to the Ki values for the interaction with the imidazoline I1 and I2 receptors are provided in Figure S1 in the Supporting Information.
compounds. A first set of 14 binding assays was selected on the basis of prior evidence of relevance to AD.18 This included the serotonin 5-HT1A, 5-HT2A, and 5-HT2C receptors, an unspecific α adrenoceptor type 2 (α2), the dopamine D2 receptor, the muscarinic M3, M4, and M5 receptors, the three δ-, κ-, and μ-opioid receptors (DOP, KOP, and MOP), the serotonin 5-HT3 ion channel, and the acetylcholinesterase (AChE) and monoamine oxidase type B (MAO-B) enzymes. In addition, similarity-based computational methods were recently applied to predict the likely protein targets of small molecules bioactive in phenotypic screens.19 Accordingly, the previous selection was complemented with the results obtained from an in silico target profiling of compounds 9, 10, and 17 using a similarity-based approach that was recently validated both retrospectively, on its ability to predict the entire experimental interaction matrix between 13 antipsychotic drugs and 34 protein targets20 and to identify cancer-relevant proteins in a target deconvolution study of selective cytotoxic compounds in tumor cells,21 and prospectively, on its capacity to identify the correct targets for all molecules contained in a biologically orphan chemical library22 and to anticipate the affinity profile of the muscle relaxant drug cyclobenzaprine.23 Compound 9 was predicted to have affinity for the melanocortin 4 receptor (MC4R), and compound 17 was predicted to have affinity for the σ-1 receptor (SIGMAR) and nischarin. The latter is likely the I1-imidazoline receptor,24 and thus, the binding assays available for the imidazoline I1 and I2 receptors, in addition to MC4R and SIGMAR, were also included in the final list of targets to screen. In total, Memantine (as reference molecule) and compounds 9, 10, and 17 were profiled in vitro against a panel of 19 radioligand binding assays. The final results of this target fishing campaign are presented in Figure 5 as average percentages of inhibition of specific binding when tested with potent standard ligands for the respective targets (see the Experimental Section).
effects. Therefore, molecules 9, 10, and 17 represent novel chemical entities with clear enhanced protective effects against neuronal apoptosis. Target Deconvolution. Because the novel neuroprotective molecules identified were originally selected from chemical providers by pharmacophoric feature similarity to NMDAR antagonists, our first assumption was to contemplate NMDAR as the obvious target for compounds 9, 10, and 17. Accordingly, these three compounds were submitted for testing on radioligand binding assays for the four binding sites (agonist/glutamate, glycine, phencyclidine, and polyamine) reported to regulate NMDAR activity. Memantine and MK801, as reference NMDAR antagonists, and compound 5, as representative of the group of compounds retaining the essential pharmacophore features of MK-801 (vide supra), were also tested on the same four assays. The results, shown in Figure 5, confirmed a full binding affinity of Memantine (92% at 10 μM) and MK-801 (98% at 10 μM) for the phencyclidine site of the NMDAR and consistently revealed significant affinity on this assay also for compound 5 (69% at 10 μM). However, most surprisingly, compounds 9, 10, and 17 were found to be completely inactive in all NMDAR assays. We had then three compounds that showed significantly better neuroprotective effects than Memantine in an Aβ-induced apoptosis phenotypic assay (Figure 3), but unlike Memantine, they were not interacting with the phencyclidine site of the NMDAR. From a puristic standpoint, because these three compounds were selected based on their similarity to Memantine and MK-801, they would be considered false positives in the virtual screen. Therefore, a target deconvolution effort to identify the potential target(s) of those compounds was initiated. For the sake of clarity, this target identification is unrelated to the virtual screen and only makes sense in this particular setting that combines in silico and in vitro approaches. To this aim, a screening panel of in vitro assays was defined to further extend the pharmacological profile of the 9842
dx.doi.org/10.1021/jm301055g | J. Med. Chem. 2012, 55, 9838−9846
Journal of Medicinal Chemistry
Article
or by other mechanisms, as allosteric modulation via nonradioliganded binding sites. Characterization of the affinity profile of compounds 9, 10, and 17 across a set of 19 AD-relevant targets revealed that their mechanisms of action are unique as compared to those established for current AD drugs and other multitarget strategies proposed recently.27−30 Essentially, apart from the low micromolar affinity identified for I2 (9, 10, and 17), I1 (9), and σ-1 (17) receptors, no significant affinity (
